Many neural circuits are formed using a two-stage developmental process that includes an initial period of exuberant synapse formation, followed by a longer period of activity-dependent synapse elimination. While pruning has been well-documented in many organisms and brain regions, little attention has been paid to the rate at which synapses are eliminated during development. Different pruning rates have a strong effect on the quality and plasticity of the resulting neural circuits, and indeed, many neurological diseases have been linked to abnormal synapse levels in the cortex during critical developmental periods. We hypothesize that pruning rates in the cortex have been optimized to achieve both connectivity and robustness of underlying neural circuits. To test our hypothesis, we propose a joint experimental-computational research plan to explore how synapse levels change during development and how these changes manifest at the network and phenotypic levels. First, we will quantify the precise rate of synapse elimination during development of the mouse barrel cortex using a specialized electron microscopy (EM) preparation that selectively stains for synapses. We will develop a fully-automated and high-throughput image processing pipeline that will allow us to count tens of thousands of synapses per time point and gain robust statistics of pruning rates and the time of peak synapse density. Second, we will develop computational models of synaptic pruning to evaluate how pruning rates affect information processing in neural circuits. Our computational experiments will explore how sparse circuits emerge that are capable of efficient and robust encoding, while still being flexible enough for plasticity and adaptation and while meeting metabolic costs. Our models will also help us answer questions about global neural circuitry, including whether hubs are likely to exist and how functional modules are formed. Third, we will examine cases of synapse rewiring and reorganization following abnormal developmental conditions. We will repeat our EM procedure using mouse models of Fragile X syndrome and Rett syndrome to compare synapse levels with respect to control, and we will extend our computational models to understand circuit-level differences in these conditions. The proposed project will generate new computational and experimental solutions for analyzing network development and will lead to biological insights into how local pruning mechanisms affect global circuit properties.